High Quality Annotations of Power Infrastructure in Rural Ontario

We then leveraged the Amazon Mechanical Turk annotation environment to obtain high confidence annotations of whether the images contain power-related infrastructure or not. This data exemplifies the joint use of S^3 and the Amazon MTurk framework for machine vision-related applications.

The images in the dataset are paired: one image is forward-facing in relation to the road while it's compliment is rear-facing. These are denoted by a unique ID followed by "A" or "B". Contents: - ontario_images.tar.gz : Compressed file containing the 18,883 images used in our analysis. - classification_results.csv : The latitude, longitude, heading, and 5-class annotation results from five Amazon MTurk workers. - mapping.csv : A mapping file relating the unique image ID to the image file name. - 5 x Example_Image : Randomly selected images from the dataset